REVIEW OF CONCEPTS OF STABILITY

Victor Szebehely, Professor

D.D. Cockrell Chair
University of Texas
Austin, Texas, USA

ABSTRACT. Concepts of stability, associated nomenclature and names of

originators are reviewed emphasizing some global aspects as well as
specific applications to dynamics and to celestial mechanics. Due to
the many definitions and interpretations used, details concerning only
a few fundamental concepts (Hill's, Liapunov's, Poincare's stability,
etc.) are offered. Short definitions and descriptions are also given
for about 50 concepts of stability in the form of a dictionary.
Several of the definitions presently in use are not unique, in fact are
often contradictory and/or repetitive. Furthermore, the credits given
to the originators are not uniform or consistent in the literature and
several cross-references seem to be necessary. The list of references
given contains some fundamental works and it does not attempt to be
complete. The entire length of this paper would have been easily taken
up by a more detailed list of references, which list would still not
have been "complete".
In the "Introduction" some historical and fundamental aspects of
stability are discussed. This is followed by the review of "Major Sta-
bility Concepts" in some detail. The "Dictionary of Stability" forms
Part 3 and "Conclusions and Future Research" are in Part 4, treating
generalizations of present techniques and outlining problems of funda-
mental importance and of inherent difficulties. Part 5 gives an abbre-
viated list of references.

I. INTRODUCTION

The subject of stability is excellently suited to a COlloquium

dedicated to the memory of Alexander von Humbolt whose fundamental idea
was that the successful integration of the various fields of natural
sciences will lead to progress in human culture. Indeed, stability is
one of the fundamental subject unifying the sciences and serving

CelestialMechanics 34 (1984) 49-64. 0008-8714/84.15

9 1984 by D. ReidelPublishing Company.
S~3 V SZEBEIAEL~/

excellently the purposes of history and Dhl~osophy.

Haglhara (1957) anticipated the purpose of this Colloquium when he
considered stability a "fasclnat~ng and difficult problem of human cul-
ture". He formulated the problem of the stablllty of the solar system
as follows: "Will the present configuration of the solar system be
preserved for some long interval of time? Will the p~anets eventually
fall into the Sun or wlll some of the planets recede gradually from the
Sun so that they no longer belong to the solar system? What is the
interval of time, at the end of which the solar system deviates from
the present configuration bY a previously assigned small amount?" Then
Haglhara continues: "The question has long been an acute problem in
celestial mechanics since Laplace, not to say the Egyptian or the Cal-
dean civilization. The term "stability" has often been discussed by
various mathematicians and the solution of the problem becomes more and
more complicated and difficult to answer as we dig deeper and deeper
into it. Present day mathematics hardly enables us to answer this
question in a satisfactory manner for the actual solar system. We must
limit ourselves here to describing the present status of the efforts
toward solving thl8 fascinating but difficult problem of human cul-
ture."
Solar system stability problems appear to form a rather small
sub-set of stability research. Indeed, Prigoglne (1980) points out
that the origin of life may be related to successive instabilities (or
bifurcations). In thermodynamics and statistical mechanics we study
the role of entropy and in modern dynamics we 0eacrlbe chaotic motion
and instability by Kolmogorov's entropy.
Moving to even more general ideas, consider the stability of
theorems in any field of science. Often theorems are formulated as
follows. "If certain conditions are given, then a specific statement
is true." We might ask by whom are these conditions given and how well
are they fulfilled in any actual application? Small errors and uncer-
tainties in the "given conditions" might destroy the validity of the
conclusion if the theorem is unstable to small disturbances. Such
theorem-stabillty might be applicable to the establishment of physlcal
laws which are often based on observations with uncertainties. What
errors are allowed (and how do we estimate the actual errors and uncer-
tainties) in order not to destroy the validity of the stated law as the
consequence of observations? Clearly, the sensitivity of laws and
theorems to observational errors or to the uncertainties of the condi-
tions is critical.
We might go one step further and state that depending on the
 REVIEW OF CONCEPTS OF STABILITY 51

specific definition of stability used, real systems may always present

instabilities when suitably large disturbances are introduced. Once
again, the important idea is to find the "proper" disturbance and the
"proper" stability condition when a given system or phenomenon is
investigated. We must realize that stability depends on the distur-
bance allowed and on the type of equivalence we want (how much devia-
tion is allowed when the system is still called "stable"). Further-
more, disturbances should be of the relevant type and physical repeata-
bility should be demanded only for striking features. This leads us to
the idea of qualitative versus quantitative stability.
We might realize that no physical laws, no theorems, no constants
and no initial conditions are exact, consequently, the results of any
investigation (philosophical, theoretical, experimental, numerical,
etc.) will depend on the stability of the problem investigated (Bell-
man, 1968 and Szebehely, 1979). The answer might not be correct if the
system is sensitive to errors made by our approximations, that is if
the system is unstable. On the other hand the theories derived, the
physical laws established and in general, our results will be "good
approximations" (to what?) if the system is stable. The above discus-
sion will alert the reader to the overriding importance of the concept
of stability and of its application to a wide variety of fields in
addition to mathematical physics, astronomy, celestial mechanics and
engineering.
The ultimate step of course is the realization that no matter what
field one pursues, stability ideas always need precise considerations
and careful follow-ups. Indeed, to accomplish any results of "vali-
dity" the stability of the approach can not be ignored since uncertain-
ties in models, theories, and in various inputs will influence the
results.

2. MAJOR STABILITY - CONCEPTS

The reader will not be surprised to find after the preceedings

that there are at least 50 terminologies and concepts of stability,
regularly used in the literature. In this Section Hill's, Liapunov's,
Poincare's and Kolmogorov's ideas are selected to offer some founda-
tions and details. As will be shown later, not all but most of the
stability concepts used are associated with these ideas. The advan-
tages and disadvantages of these methods depend of course on the appli-
cations, nevertheless, s o m e g e n e r a l ideas and fundamental differences
will be outlined.
52 V SZEBEHELY

P.1 Hill's Stability

Hlll (1878) proposed the use of the energy-lntegral to establish

bounds of motlon for conservatlve systems with tlme-lndependent poten-
tlals. In general and slmpllfled terms, the basic idea may be stated
as follows. Let

v 2 = 2F(ql,q2 ... qn ) ~ C, (I)

where v is the velocity (v 2 - 61~i, using the summation eonvention), F

is the potential funetlonjq i are the coordinates and C is a constant of
the motion. Note that Hill's original idea was to establ~sh the boun-
darles of the planar lunar orbit, using his slmplifled potential and
the associated Jaeoblan integral. The method might be used for any
problem where IntegralCs) of the motion exist. As long as the degree
of freedom Is less than three the associated Hill surfaces might be
constructed and the possible and forbidden regions of the motion mlght
be established. Note that the considerable advantage of Hill's method
is that knowledge of solutions are not needed to estaDllsh stability, a
slEnlficant fact when non-integrable dynamical systems are investigated
as is usually the case in practically important eases, On the other
hand it is of considerable importance to give the proper interpreta-
tlons of the results obtained by Hill's method. For instance, in the
lunar problem Hill's method shows that the Moon's orbit Is bounded
around the Earth, using the approx[matlons of the circular, planar res-
trlcted problem of three bodies (Sun-Earth-Moon). It Is interesting to
note that Hill's simplified version of the lunar theory shows higher
stability for the Moon than using the model of the complete restricted
problem. And if the "restricted" assumption gives way to the use of
the general problem of three bodies then the generalization of Hill's
method shows Instabillty. These remarks demonstrate that the Moon's
stability according to Hill's method shows model-dependence, a not
unexpected but at the same time not generally known result.
Many additional comments could be made about Hill's method but for
brevlty's sake only a recent result is mentioned. A quantitative
measure of stability replaces Hill's originally qualitative method
(Szebehely 1977), as follows. If the constant of integration of the
orbit to be [nvestlgated is Cac (for actual) and the critical value (at
which bifurcation occurs} of the same constant is Cot , then the measure
of stability is defined by S ~ (Cae - Ccr)Cc-I
r. (Note that C cr ~ 3 in
t~e restricted problem using proper deflnltlons and units.) The
 REVIEW OF CONCEPTS OF STABILITY 53

physical meaning of this measure is that it shows the distance in phase

space between the actual and the critical orbits. If S < 0 we have
Hill-type instability, suggesting that the orbit may be unstable and,
in the case of the lunar orbit, the Moon may leave the Earth and become
a planet of the Sun. The dimensionless generalization of the Jacobian
constant (C) in the case of the general problem of three bodies (Zare,
1977) is C = - hc2/ ~5G2, where h is the total energy, c is the angular
momentum, ~ is the average mass of the participating bodies and G is
the gravitational constant. The stability measure is given by the same
formula as before but the limiting surfaces never close completely
inspite of the use of the energy-integral and of the momentum integral
because of the high degree of freedom (9 vs 3).
9 l
2.2 Definitions of Liapunov's and of Polncare's Stability

Liapunov and Poincare'contributed several fundamental ideas to the

field of stability. In this section a comparison of their considerably
similar basic ideas is offered and in the next sections Liapunov's
9 l
characteristic number, Liapunov's function, Po~ncare's surface of sec-
tion and Kolmogorov's representation on tori are discussed.
The two fundamental concepts contributed by Liapunov (1892) and
t
Poincare (1892) differ in their selection of correspondence between
points on the original orbit (the stability of which is to be investi-
gated) and points on the disturbed orbit. Liapunov established "iso-
chronous" correspondence, while Poincare referred to "normal"
correspondence. Referring to Figure I, we show the original orbit (A)
and the disturbed orbit (B). The normal correspondence is represented
by the deviation ~ and the isochronous relation by ~.
The analytical expression and the definition of Liapunov's stabil-
ity are shown first.
Given the system of differential equations for the vector ~ as
x = F(~,t) with a solution x = ~(t) as shown by the curve A on Fig. I.
This solution is stable if given any ~ > O, there exists a 6 > 0 such
that

I[oI = l (to ) - (to)l <

IYl = I T ( t ) - ~(t)l < ~-

Here ~(t) is the solution obtained when t h e disturbance is applied,

shown as curve B. As s e e n from the inequalities as well as from Figure
54 V SZEBEHELY

t=t! B

t=t
Yo ~ ~ t =to

Figure i. Isochronous and Normal

Correspondences

pl~ B ~

P O ' ~ A
,oo.~~~,
Figure 2. Computation of Liapunov's
Characteristic Exponents
 REVIEW OFCONCEPTSOFSTABILITY 55

1, Liapunov's stability refers to isochronous correspondence.

Using the normal vector ~ instead of y we obtain Poincar~'s "nor-
mal" stability also known as orbital stability.
If either ~ or ~ goes to zero as t + ~ we speak about asymptotib
stability. Note that when the semi-major axis (a) of an elliptic two-
body orbit is disturbed (changed) by ~ , the associated e value will
become 2a after sufficient number of revolutions. Consequently,
elliptic orbits are, according to Liapunov's definition unstable but
orbitally stable.
The often controversial idea of influencing stability by transfor-
mations should be mentioned here. If stability is an invariant pro-
perty of a dynamical system, then transformations of the variables
should not affect the stability properties. Consider for instance
Levi-Civita's (1903) transformation or its generalization, the
Kustaanheimo-Stiefel (1971) transformation. The unstable elliptic
two-body orbit appears now as the stable solution of a linear differen-
tial equation representing a harmonic oscillator. In this differential
equation, however, the energy appears as a constant, therefore, no per-
turbations are possible which would change the energy and this fact
offers Liapunov stability. The situation is similar when the equations
of motion are formulated using orbital elements.

2.3 Liapunov's Characteristic Numbers and the Liapunov Functions

Liapunov's (1892) characteristic numbers indicate the stability of

solutions by means of applying repeated disturbances and n u m e r i c a l l y
integrating the outcome. Figure 2 shows the process. Curve A is once
again the original undisturbed solution and the segments BI,B2 ' etc.
represent the disturbed solutions.
The change of one of the initial conditions at t = t o is d o so the
integration of the disturbed orbit starts at P o" When point P'I is
reached the original disturbance, d o = PiP1 '' is once again introduced
and the segment B 2 is obtained. Consequently if d i = PiP'i and
do = Po P o = PIPI = PiPi" we have

di Pi P i
do po P -o 9

Liapunov's characteristic number is defined by

L = lim L
n
56 V SZEBEHELY

I dl
where L n - ~-~ E in ~--
n=T O

If L > 0 we have instability or chaotic mot/on, if L - O we have

quasi-periodlc motion and if L < 0 we have asymptotic stability or
attraction.
Note that the solution must be known for the application of this
approach as well as repeated accurate numerical integrations are
needed. This might be considered a disadvantage of thim method when
compared with Hill's approach. On the other hand Liapunov's charac-
teristic number will certainly describe the details of the stability
aspects of a given solution for various disturbances while Hill's
method will offer only possible boundedness without consideration of
t~e specific orbits.
Liapunov's theorem (almo known as his direct method) does not
require the knowledge of a solution and offers the stability informa-
tion in a region of the phase-space by means of establishing the
existence of Liapunov's function. The L-function associated with the
differential equation x i = XI(Xj) Is V(xj) if

V dV X 3V
= :~-" i~-~ ~ 0

The existence of the L-function offers stabllity, in fact we have

asymptotic stability if -V is positive definite. If V is po~Itlve
definite, we have instability.

2.4 Polnear~'s Surface of Section

One of the important method in modern dynamics is the numerical

use of Poincare's (1892) surface of section technique which reveals the
phase space behavior of two-degrees of freedom dynamical systems, From
this, important conc~uslons might be drawn concerning stability. The
technique became widely used when hlgh-speed computations reached suf-
ficient level concerning accuracy and speed. While it Is applicable to
higher than two-degrees of freedom dynamical systems, most users limit
it to the study of ~uch systems. The basic idea might be described by
considering the system ~ - F(x,y,x,y,t) and ~ - O(x,y,x,y,t).
After numerical integration we prepare a two-dlmenslonal represen-
tation by plotting the series of points P(y,y) corresponding to
x - 0 and x > 0. In other words, the four-dimenslonal flow in the x,
x, y, y phase-space is represented by the points of intersections of
 REVIEW O F CONCEPTS O F S T A B I L I T Y 57

the trajectory with the x = 0 plane. This might also be looked at as a

point-transformation in the y, ~ plane, the transformation resulting in
the intersections. If the transformation has an invariant point, this
will correspond to a periodic orbit, while smooth curves in the y,
plane represent quasi-periodic solutions. If the points are covering
the plane we speak about random or chaotic motion. Application to sta-
bility now consists of establishing such plots and observing the possi-
ble change of the qualitative nature of the solutions as disturbances
are applied.

2.5 Kolmogorov's Tori

Kolmogorov's original idea proposed in 1954 is now known with

various proofs, several modifications and improvements as the
Kolmogorov-Arnold-Moser (KAM) theorem (Arnold 1963, Moser 1966). The
basic idea is that an unperturbed two-body motion can be represented as
a flow on a torus. Will the motion stay on a distorted torus when per-
turbations, resonances, etc. are applied or will the torus completely
disappear? Will the flow cover the torus completely (corresponding to
quasi-periodic orbits with irrational commensurabilities), will the
flow close after a finite number of revolutions (corresponding to
periodic orbits) or will the resonances combined with the perturbations
result in the elimination of the torus, corresponding to chaotic
motion? The KAM theory offers conditions for pertubations and commen-
surabilities, which if satisfied will preserve the torus. Once again
stability is associated with qualitative changes of the trajectories.
When some orbits in multi-dimensional space drift between KAM tori we
speak about Arnold's diffusion. The result is emerging that non-
integrable systems have non-zero measure of unstable regions of the
phase space in general.

3. DICTIONARY OF STABILITY

In this section short definitions are given of various concepts

and terminologies used in stability research. The definitions will not
be given by mathematical relations for the sake of brevity, especially,
since the basic concepts discussed in Section 2 will be frequently
referred to. We will use the abbreviation st for stability and will
try to avoid other, possibly confusing abbreviations. Only few of the
fundamental concepts of linear stability analysis are included. Note
that some of the definitions are repetitive, some contradictory,
nevertheless, they represent the present state of art.
fi$ v SZEBEHELY

asymptotic st: effect of disturbance as t ~ ~; convergent st; extreme

st

characteristic roots: solutions of the characteristic equation; elgen-

values of the matrix representing system of linear equations

characteristic st: st of a system without perturbations

complete st: asymptotic st in the large; st to all orders; practical

st; formal st

conditional st: st of a subset of solutions

convergent st: see asymptotic st

discrete st: st conditions satisfied for certain discrete times

dangerous reglon: region of instability

extreme st: see asymptotic st

first order st: see linear st'

formal st: see complete st

future st: st for 0 S t ~ § |

hlerarchlcal st: no orbit crossings In planetary systems; no change of

the order of the Jaeoblan distances

Hill st: solutions bounded in the configuration space

infinitesimal st: see linear st

lsoehronous st: see text

L-st: Liapunov st; strong st; see Isoehronous st

Lagrange st: st of equilibrium solutions; all solutions stay in bounded

regions of the phase space

Laplace st: all solutions stay in bounded regions of the phase space;
no collisions and no escapes for many-body problems

Llapunov st: see isoohronous st;

linear ~t: st of linear or linearlzed systems;

Lindblad st: st of stars at edges of flat nebulae

 REVIEW OFCONCEPTSOFSTABILITY 59

Malkin st: st under persistent perturbations; modified L-st

negative st: st for 0 ~ t § -

neutral st: st of equilibrium when other equilibria are near by

non-linear st: st of non-linear systems

normal st: see text

numerical st: the st of the method used for numerical integration

orbital st: see text

permanent st: see future st

planetary system st: no secular terms in the semi major axes

/
Poincare st: see text

Poisson's st: system returns infinitely often to positions arbitrary

near to its original position

positive st: see future st

practical st: magnitude of disturbances related to characteristic

roots; uniformly bounded solution for finite time; effect of ini-
tial conditions and disturbances;

S-st: temporary st for 0 ~ t ~ S

safe region: region of st

secular st: no secular terms appear in the solution

semi-permanent st: temporary stability for long time

straight-forward st: effect of initial conditions

structural st: effect of change in model, chage of constants or change

of form of the differential equations

temporary st: see st for finite time

topological st: topological equivalence of the set of disturbed and

original solution curves

trigonometric st: no real parts of the characteristic roots and no

commensurability for linear systems
60 v ~ZEBEMEI~Y

unllateral st: asymptotic st without oscillation; only negative real

parts of the characteristic roots

vertical st: st in the direction normal to the plane of the orbits of

the primaries in the restricted problem of three bodles

weak st: two different kinds of motion in the same neighborhood in the
phase space.

4. CONCLUSIONS AND FUTURE RESEARCH

It is hoped that the reader at the end of thls revlew arrives at

the conclusion that stability research is one of the most challenging
undertaking for the human intellect. Indeed, it covers much wlder
fields than orbit mechanics and it extends from the predictions of the
future behavior of all phenomena to the establlshment and val~dlty of
physical laws, theorles and theorems. Instabilities might be respon-
sible for the initiation of our existence or for its termination, as
mentioned in the introduction.
Regarding future stability research, it should not be surprlzlng
to know that its direction and success are highly unpredictable. Even
wlth precise definitions of stability and assuming that our present
state is well known, the outcome should be rather uncertain, according
to the basic principles mentioned in this paper. Nevertheless, it
might be expected that as new engineering systems are developed or new
physical concepts emerge, they wlll he subjected to stability analysis.
It is now clear that no single stabillty technique or definition is to
be applied for different phenomena. Flexibility Is expected and recom-
mended concerning the selection of the method of analysis to best suit
the specific system and the special dominant questions of importance.
The basic concepts (Hill's, Kolmogorov's Liapunov's, Polncare's, etc.)
will dominate stability research for the time being but new concepts
should emerge, dictated by the practical requirements of various new
systems.
The behavior of systems In the phase-space might show qualitative
changes as the system-parameters change (structural stability) or as
the initial conditions vary (straight forward stability). The essen-
tlally new idea is emerging according to which the qualitative changes
are to be emphasized s~nce small changes not effeeting the general
behavior of the system might be ignored. The change from well ordered
and well predictable periodic or quasl-perlodlc motion to chaotic
behavior is to be predicted or to evaluated. Such changes are not
expected for linear systems, therefore, interest is directed to systems
 REVIEW OF CONCEPTS OF STABILITY 61

governed by non-linear laws.

A recent example, subject to further research might be shortly
described here to illustrate the point. The motion about the triangu-
lar equilibrium points of the restricted problem of three bodies is
considered stable if the value of the mass- parameter (U) is suffi-
ciently small. This condition is usually satisfied in the solar system
considering various combinations of planets and their satellites. If
the particle is not at the equilibrium point initiaily, i.e. if the
initial conditions deviate by a "small" amount in the phase space, the
motion is known as "libration" around the equilibrium point. As the
initial conditions are varied the limit of the librational region is
reached and "circulation" sets in. (The phenomenon shows some analogy
to the motion of the spherical pendulum.) Further increase of the devi-
ations results in chaotic motion, escape, capture, etc. The product of
the limiting deviations in the configuration space (A qi ) and in the
momentum space (A pi), according to some of our preliminary results
(Szebehely, 1979) give a constant order of magnitude, showing a
strictly formal analogy to Heisenberg's uncertainty relation. In other
words,

3
A qi A Pi = K = constant,
i=I

regardless of the direction along which the deviations were applied.

Large limiting initial deviations in the position are associated with
small limiting initial deviations in the momentum in any given direc-
tion {and vice versa). The value of K gives the combination of the
limiting deviations which results in non-librational or chaotic motion.
If the value of the product is less than K, the motion is libration
(periodic or quasi-periodic). If the value of the product is larger
than K, the motion is circulation, capture, escape or chaotic. The
motion changes qualitatively as the product exceeds the value of K~ In
this way the behavior depending on the initial conditions is evaluated.
Since the critical value of K is expected to change with the mass-
parameter, the dependence of the behavior on the system-parameter, ~ is
to be evaluated. In other words, straight forward and structural sta-
bility are evaluated corresponding to change of the character of the
motion.
The above example might indicate the future direction of stability
research of dynamical systems. Stability might be associated with
change of the values of the initial conditions or with change of values
62 V. SZEBEHELY

of the system-parameters. The region of phase space for which the type
of motion is the same represent a stable region. The region where the
motion changes qualitatively might be termed unstable.

Acknowledgements

This paper could not have been prepared without the contributions
and cooperation of my colleagues, co-workers and students who influ-
enced my work on stability for the last thirty years. Support received
from the National Science Foundation and from the University of Texas
to prepare this paper and to attend the Alexander von Humbolt Confer-
ence is gratefully acknowledged.

REFERENCES

The following list concentrates on the fundamental and basic

reference articles in addition to recent books which the reader might
find useful. Note that some of the text book references of this list
are not mentioned in the present paper and they cover many asPects of
stability of rather general interest.

Arnold, V.I., Dokl. Akad. Nauk, SSR 137, 255, 1961; 138, 13, 1961; 142,
758, 1962; 145, 487, 1962. Also Usp. Mat. Nauk, 18, 13, 1963;
18, 91, 1963.

Bellman, R., "Some Vistas of Modern Mathematics", Univ. of Kentucky

Press, 1968.

Birkhoff, G.D., "Collected Mathematical Papers", American Math. Soc.,

N.Y., 1950.

Brillouin, L., "Scientific Uncertainty and Information", Academic

Press, N.Y., 1964.

Chandrasekhar, S., "Principles of Stellar Dynamics", Dover Publ., N.Y.,

1960.

Hagihara, V., "Stability in Celestial Mechanics", Kasai Publ., Tokyo,

1957.

Helleman, R.H.G., in "Fundamental Problems in Statistical Mechanics",

E. Cohen, Editor, p. 165, North Holland Publ., Amsterdam, 1980.
 REVIEW OF CONCEPTS OF STABILITY 63

HiI1, G., "Collected Mathematical Works", Carnegie Inst., Washington,

D.C., 1905-1907.

Horton, C.W., L.E. Reichl and V. Szebehely (Editors), "Long-Time Pred-

iction in Dynamics", Wiley Publ., 1963.

Jefferys, W.H. and V. Szebehely, Comments of Astrophysics, 8, 9, 1978.

Jorna, S. (Editor), "Topics in Nonlinear Dynamics", American Inst. of

Physics, N.Y., 1978.

Kolmogorov, A.N., Proc. Intern. Cong. Math., Amsterdam 1954, ~, 315,

North Holland Publ. Amsterdam, 1957 and Dokl. Akad. Nauk, SSR
98, 527, 1954.

LaSalle, J. and S. Lefschetz, "Stability by Liapunov's Direct Method

with Applications", Academic Press, 1961.

Levi-Civita, T., Ann. Math. [3] 9 I, 1903.

Liapunov, A.A., Communications Math. Soc. Krakow, ~, I, 1892.

Malkin, I.G. "Theory of Stability of Motion", State Publ. House,

Moscow-Leningrad, 1952.

Moser, J., SIAM Review, 8, 145, 1966.

Moser, J., "Lectures on Hamiltonian Systems", Memoirs, American Math.

Soc., 81, I, 1968.

Moser, J., "Stable and Random Motion in Dynamical Systems", Princeton

Univ., Press, Princeton, N.J., 1973.

Pars, L.A., "A Treatise on Analytical Dynamics", Heineman Press, Lon-

don, 1965.
l
Poincare, H., "Les Methodes Nouvelles de la Mecanique Celeste,"
Gauthier-Villars, Paris, 1892-1899.

Prlgogine, I., "From Being to Becoming", W.H. Freeman and Company, San
Francisco, 1980.
64 V. SZEBEHELY

Prigogine, I. and R. Herman, "Kinetic Theory of Vehicular Traffic",

American Elsivier Publishing Company, Inc., New York, 1971.

Roy, A.E., "Orbital Motion", Adam Hilger Ltd. Publ., Bristol, 1978.

Siegel, C. and J. Moser, Lectures on Celestial Mechanics, Springer,

N.Y., 1971.

Stiefel, E.L. and G. Scheifele, "Linear and Regular Celestial Mechan-

ics", Springer, N.Y., 1971.

Szebehely, V., Celestial Mechanics, ~, 116, 1971; 9, 359, 1974; 15,

107, 1977; 18, 383 and 391, 1978; 22, 7, 1980; 23, 3, 1981.

Szebehely, V., "Theory of Orbits", Academic Press, N.Y., 1967.

Szebehely, V., Proc. Nat. Acad. Sc., 75, 5743, 1978.

Szebehely, V. (Editor), "Instabilities in Dynamical Systems", D. Reidel

Publ., Holland, 1979.

Ulam, S.M., "Problems in Modern Mathematics", Interscience Publishers,

New York, 1980.

Whittaker, E.T., "A Treatise on the Analytical Dynamics of Particles

and Rigid Bodies", Cambridge Univ. Press, London, 1904.

Wintner, A., "Analytical Foundations of Celestial Mechanics", Princeton

Univ. Press, Princeton, 1947.

Zare, K., Celestial Mechanics, 16, 35, 1977.